If You Look at ... Lexical Bundles in University Teaching and Textbooks PDF

Applied Linguistics 25/3: 371405 # Oxford University Press 2004
If you look at . . .: Lexical Bundles in

University Teaching and Textbooks
DOUGLAS BIBER1, SUSAN CONRAD2, and VIVIANA CORTES3
1
Northern Arizona University, 2Portland State University, and 3Iowa State University
This paper investigates the use of multi-word sequences in two important

university registers: classroom teaching and textbooks. Following Biber et al.
(1999), we take a frequency-driven approach to the identication of multi-
word sequences, referred to as `lexical bundles'. We compare the lexical bundles
in classroom teaching and textbooks to those found in our previous research on
conversation and academic prose. Structural patterns are described rst, and
then we present a functional taxonomy, including stance expressions, discourse
organizers, and referential expressions. The use of lexical bundles in classroom
teaching turns out to be especially surprising, both in frequency and in
function. Classroom teaching uses more stance and discourse organizing
bundles than conversation does, but at the same time, classroom teaching
uses more referential bundles than academic prose. The analysis indicates that
lexical bundlesthe most frequent sequences of words in a registerare a
unique linguistic construct. Lexical bundles are usually not complete
grammatical structures nor are they idiomatic, but they function as basic
building blocks of discourse. In the conclusion, we discuss the implications of
our study for the theoretical status of lexical bundles.
1 INTRODUCTION
There have been many linguistic studies of academic discourse published over
the past 20 years (e.g. the extensive survey of research in Grabe and Kaplan
1996). Most of these focus on the description of a specic grammatical feature
or lexical class in written academic registers, especially academic research
articles in science or medicine (e.g. Halliday 1988; Myers 1989; Swales 1990;
Thompson and Ye 1991; Halliday and Martin 1993; Hyland 1994, 1996a,
1996b; Williams 1996; Crompton 1997; Grabe and Kaplan 1997; Swales et al.
1998; Chih-Hua 1999; Marco 2000; see also the survey of grammatical
characteristics in Biber et al. 1999).
Relatively few studies describe the linguistic characteristics of spoken
academic discourse. However, some of these studies take a dierent approach,
focusing on discourse markers and other relatively xed lexical `chunks'
instead of grammatical features (e.g. Chaudron and Richards 1986; Strodt-
Lopez 1991; Nattinger and DeCarrico 1992; Flowerdew and Tauroza 1995).
These studies of discourse markers and other lexical expressions in
academic lectures are part of a growing research tradition focusing on the
372 LEXICAL BUNDLES IN UNIVERSITY TEACHING AND TEXTBOOKS
use of multi-word prefabricated expressions (e.g. in a nutshell, if you see what I

mean). This research challenges the view that language is strictly composi-
tional, arguing instead that much of our everyday language use is composed
of prefabricated expressions (see the reviews in Ellis 1996; Howarth 1996;
Wray and Perkins 2000; Wray 2002).
Multi-word sequences have been studied under many rubrics, including
`lexical phrases', `formulas', `routines', `xed expressions', `prefabricated
patterns' (or `prefabs'), and `lexical bundles'. Many previous studies have
been primarily theoretical in nature, comparing the various perspectives and
approaches to multi-word units, proposing new frameworks for analysis, and
calling for further research. Hakuta (1974), Yorio (1980), Pawley and Syder
(1983), Redeker (1991), Sinclair (1991), Lewis (1993), Weinert (1995),
Howarth (1996, 1998a, 1998b), and Wray and Perkins (2000) are good
examples of this type. Weinert (1995: 182) identies two basic research issues:
the best way to dene and identify xed multi-word units, and analysis of the
discourse functions that these multi-word units perform.
Numerous empirical studies have addressed aspects of these two research
issues (e.g. Renouf and Sinclair 1991; Kjellmer 1991; Nattinger and
DeCarrico 1992; Altenberg and Eeg-Olofsson 1990; Altenberg 1998; Aijmer
1996; Francis et al. 1996, 1997, 1998; Butler 1997, 1998; Hudson 1998;
DeCock 1998; Granger 1998; Howarth 1998a, 1998b; Moon 1998a, 1998b;
Partington 1998; Hunston and Francis 1998, 1999; Gledhill 2000; Schmitt
2004). However, despite the general consensus on the importance of multi-
word units, there is surprisingly little agreement on their dening
characteristics, the methodologies to identify them, or even what to call
them; and, as a result, there is little agreement across studies on the specic
set of multi-word units worthy of description. These empirical research
studies dier in terms of:
1 the research goals adopted: describing the full range of multi-word
sequences vs. describing a small set of ìmportant' sequences;
2 the criteria used to identify multi-word units: perceptual salience,
frequency criteria, or other;
3 the formal characteristics of the multi-word units studied: continuous
sequences, discontinuous frames, or lexico-grammatical patterns; two-
word collocations or longer sequences;
4 the text samples drawn on: ranging from a few texts to very large corpora
(c.100 million words);
5 whether or not register comparisons are made: many studies disregard
register completely; others analyse only spoken texts or written texts; a
few studies explicitly compare multi-word units across dierent registers.
Given the complexity of these issues, we take the position that no single
approach can provide the whole story. Rather, the overall importance of
multi-word units in discourse can be fully understood only by undertaking
empirical research studies from dierent perspectives. In the present paper,
DOUGLAS BIBER, SUSAN CONRAD, and VIVIANA CORTES 373
we adopt a frequency perspective`lexical bundles'to investigate the

functions of multi-word sequences in university registers.
1.1 Previous research on lexical bundles

The present study adopts a frequency-driven approach to multi-word units,
based on analysis of the most frequent recurrent sequences of words. Salem
(1987) carried out pioneering research of this type, based on analysis of a
corpus of French government documents. Altenberg (1998; Altenberg and
Eeg-Olofsson 1990) was probably the rst researcher to study recurrent word
sequences in English (based on the London-Lund Corpus), while Butler
(1997) adopted a similar approach for his analysis of recurrent word
sequences in a large corpus of Spanish texts. We extended this research
approach in the Longman Grammar of Spoken and Written English (Biber et al.
1999, ch. 13; see also Biber and Conrad 1999), referring to these recurrent
sequences of words as `lexical bundles' (see section 2 below). That study used
corpus-based research methods to compare the most common multi-word
units in spoken and written registers. That research was distinctive in several
respects:
. it adopted a register perspective and explicitly compared spoken and
written registers (conversation and academic prose);
. it was based on empirical analysis of large corpora (c.5 million words for
each register);
. it relied exclusively on frequency criteria for the identication of multi-
word units;
. it focused on longer multi-word units than in most previous studies: 4, 5,
and 6-word sequences.
In a subsequent study, we developed a preliminary taxonomy to classify the
discourse functions of the lexical bundles found in conversation and academic
prose (Biber et al. 2003; Conrad and Biber, in press). This analytical
framework has also been applied to other more specialized domains. For
example, Cortes (2002, to appear) applies the lexical bundle framework to the
study of university student writing in history and biology, comparing the use
of lexical bundles by professional and student authors. Partington and Morley
(2004) investigate the use of lexical bundles in spoken political discourse
(White House press briengs and political news interviews), showing how
lexical bundles can be used to analyse the distinctiveness of particular
registers.
1.2 Overview of the present study

In the present paper we turn to the use of lexical bundles in university-
level courses. Specically, we analyse the use of lexical bundles in
university classroom teaching and textbooks. Classroom teaching and
textbooks are arguably the two most important registers in the academic
lives of university students. However, we know surprisingly little about the
language of these registers. To our knowledge, only one previous study has
investigated the use of multi-word units in university lectures: the study by
DeCarrico and Nattinger (1988; see also Nattinger and DeCarrico 1992) on
`lexical phrases'. The present study extends the research ndings of
Nattinger and DeCarrico by analysing the patterns of use in a large
corpus and employing frequency criteria rather than perceptual salience as
the basis for analysis.
We compare the patterns of use in classroom teaching and textbooks to
those found in conversation and academic prose. Conversation can be
regarded as a stereotypically òral' register, characterized by high interaction,
expression of personal stance, and real-time production circumstances (see
Biber 1995). In contrast, academic prose can be regarded as a stereotypically
`literate' register, characterized by an informational rather than personal
focus, and extensive opportunity for crafting, revising, and editing the written
text. As discussed in our previous work (e.g. Biber et al. 2002, 2004),
classroom teaching and textbooks are intermediate registers with respect to
these characteristics. Classroom teaching is a spoken register, constrained by
real-time production circumstances, and marked to some extent by speakers'
personal concerns and interactions among participants. At the same time,
classroom teaching has a primary informational focus, and instructors
normally pre-plan the content and structure of their class sessions to achieve
their informational goals. As a result, we would expect that classroom
discourse would be intermediate between conversation and academic prose in
its use of lexical bundles. Textbooks are more similar to academic prose in
purpose and production circumstances, but the material must be presented in
a way that is accessible to students. Given recent trends to make textbooks
more engaging for students, we predicted that they would also be
intermediate between conversation and academic prose in their use of lexical
bundles.
In section 2 below, we describe the methodology used to identify lexical
bundles. Then, in section 3, we briey summarize some of our earlier
grammatical research on university registers. This research provides important
background to the present study, because it raises expectations concerning the
linguistic dierences among spoken and written university registers; it turns
out that the patterns of use for lexical bundles are strikingly dierent from the
patterns observed for other grammatical features.
In sections 4 and 5, we turn to the analysis of lexical bundles in university
registers. In section 4, we summarize our distributional ndings, describing
the structural types of lexical bundles found in classroom teaching and
textbooks, and comparing those to the common lexical bundles found in
conversation and professional academic prose. Then, in section 5 we
introduce a functional taxonomy of the lexical bundles in classroom teaching
and textbooks, grouping bundles into categories that serve related discourse
functions, and contrasting the use of these functional groups across spoken
and written university registers.
2 METHODOLOGY
2.1 Corpus used for the study
The present study is based on an analysis of texts from university classroom
teaching and textbooks in the T2K-SWAL Corpus (TOEFL 2000 Spoken and
Written Academic Language Corpus; see Biber et al. 2002, 2004). The T2K-
SWAL Corpus was designed to represent the range of spoken and written
registers that university students encounter in the U.S.1 The register categories
chosen for the corpus are sampled from across a large range of spoken and
written activities associated with academic life, including classroom teaching,
oce hours, study groups, on-campus service encounters, textbooks, course
packs, and other written materials (e.g. university catalogues, brochures). The
study reported here analyses only two of those registers: classroom teaching
and textbooks. Table 1 shows the composition of this sub-corpus by register
category.
Texts from classroom teaching and textbooks were sampled from six major
academic disciplines (Business, Education, Engineering, Humanities, Natural
Science, and Social Science) and three levels of education (lower division
undergraduate, upper division undergraduate, and graduate). Texts were
collected at four academic sites (Northern Arizona University, Iowa State
University, California State University at Sacramento, and Georgia State
University); the four universities are situated in dierent regions of the USA
and have dierent institutional proles. Although this sampling does not
achieve complete demographic/institutional representativeness, it does avoid
obvious skewing for these factors.
To provide a baseline for the analysis of university registers, we compare the
patterns of use to our earlier descriptions of lexical bundles in conversation
and academic prose, based on analysis of the Longman Spoken and Written
English Corpus (c.4 million words of British English conversation; c.3 million
words of American English conversation; c.5.3 million words of academic
Table 1: Composition of the sub-corpus used in the analysis

Register No. of texts No. of words
Classroom teaching 176 1,248,800

Textbooks 87 760,600
Total 263 2,009,400
Source: The TOEFL 2000 Spoken and Writing Academic LanguageT2K-SWAL

Corpus
prose; see Biber et al. 1999, ch. 1). The academic prose corpus comprises both
academic research articles (c.2.7 million words) and advanced academic books
(c.2.6 million words; see Biber et al. 1999: 324). While some advanced
academic books can also be used as textbooks, especially in graduate courses,
the corpora of academic prose and textbooks were sampled independently:
textbooks are mostly written specically for students, while the articles and
books included in the academic prose corpus are mostly written for other
professionals.
Identication of lexical bundles

Lexical bundles are dened simply as the most frequent recurring lexical
sequences in a register (Biber et al. 1999, ch. 13). In the analysis of lexical
bundles, as in our previous corpus-based studies of grammatical patterns, we
do not regard frequency data as explanatory. In fact we would argue for the
opposite: frequency data identies patterns that must be explained. The
usefulness of frequency data (and corpus analysis generally) is that it
identies patterns of use that otherwise often go unnoticed by researchers.
Frequency data have additional importance for the study of multi-word
sequences because they are one reection of the extent to which a sequence
of words is stored and used as a prefabricated chunk, with higher frequency
sequences more likely to be stored as unanalysed chunks than lower
frequency sequences. (Of course, frequency is only one measure of the
extent to which a multi-word sequence is prefabricated; sequences with
idiomatic meanings are usually rare but clearly prefabricated.)
The actual frequency cut-o used to identify lexical bundles is somewhat
arbitrary. For the present study, we take a conservative approach, setting a
relatively high frequency cut-o of 40 times per million words to be included
in the analysis. Many of the bundles described here are actually much more
common, occurring more than 200 times per million words. (In contrast, we
included word sequences that occurred only 10 times per million words in the
Biber et al. (1999) study.) To further limit the scope of the investigation here,
only four-word sequences are considered; for example, do you want to and I
don't know what are common four-word lexical bundles in conversation. (The
text excerpts in sections 4 and 5 below show that two four-word lexical
bundles sometimes occur together to form a ve-word or six-word sequence;
see Biber et al. (1999: 992 ) for discussion of these longer lexical bundles.)2
A sequence must be used in at least ve dierent texts to be counted as a
lexical bundle, to guard against idiosyncratic uses by individual speakers or
authors. In practice, this restriction has little eect, because most bundles are
distributed widely across the texts in a corpus. Even the least common lexical
bundles in the present analysis (with a frequency of 40 per million words) are
used in at least 20 dierent texts, while the more common bundles are
distributed more widely.3
In general, most lexical bundles are not idiomatic in meaning. In fact, most
longer idioms are far too rare to be considered bundles. Stereotypical idioms
such as kick the bucket (meaning `die') and a slap in the face (meaning àn
aront') are rarely attested in natural speech or writing. When idioms and
xed formulas are used, they occur usually in ction rather than in actual
face-to-face conversation. For example, kick the bucket and a slap in the face
occur around 5 times per million words in ction, while they are rarely
attested at all in actual face-to-face conversation (see Biber et al. 1999:
10246).4
Similarly, most lexical bundles do not represent a complete structural unit.
For example, only 15 per cent of the lexical bundles in conversation can be
regarded as complete phrases or clauses, while less than 5 per cent of the
lexical bundles in academic prose represent complete structural units (see
Biber et al. 1999: 9931000). Instead, most lexical bundles bridge two
structural units: they begin at a clause or phrase boundary, but the last
words of the bundle are the rst elements of a second structural unit. Most of
the bundles in conversation bridge two clauses (e.g. I want to know, well that's
what I), while bundles in academic prose usually bridge two phrases (e.g. in
the case of, the base of the).
Our research approach to the study of lexical bundles is deliberately
exploratory. We start out by simply asking whether there are chunks of
languagesequences of wordsthat are used repeatedly by speakers and
writers. The answer to this question is `yes': there are many lexical bundles
used with high frequency in texts, and it further turns out that dierent
registers tend to rely on dierent sets of lexical bundles. These distributional
facts raise a second set of research questions: what are these word chunks,
what are their structural and functional characteristics, and how can we
explain the repeated use of these extended word chunks? For the most part,
linguists have not noticed these high frequency multi-word sequences,
probably because most previous research has focused on grammatical phrases
and clauses, disregarding the possibility of lexical units that cut across
grammatical structures. However, we show below that lexical bundles have
identiable discourse functions, suggesting that they are an important part of
the communicative repertoire of speakers and writers, even though they do
not correspond to the well-formed structures traditionally recognized by
linguistic research. We return to discussion of the theoretical status of lexical
bundles in section 6.
3 BACKGROUND: GRAMMATICAL DIFFERENCES AMONG

SPOKEN AND WRITTEN UNIVERSITY REGISTERS
In terms of their grammatical characteristics, spoken university registers are
consistently dierent from written registers (see Biber et al. 2002, 2004). Two
general patterns are found: (1) spoken registers as a group are dierent from
written registers, but (2) there are small dierences among spoken registers,
forming a cline from the more conversational registers to the more

informational registers like classroom teaching.
Figure 1 illustrates these patterns, plotting the use of nouns, verbs, and
personal pronouns in the four registers considered in the present study:
conversation, classroom teaching, textbooks, and academic prose.5 The gure
shows a major dierence between spoken and written registers in the use of
these features: nouns are much more common in the written registers than
spoken registers, while verbs and pronouns show the opposite distribution,
being much more common in the spoken registers. Nouns are slightly more
common in classroom teaching than in conversation, reecting to some
extent the primary informational purposes of teaching in contrast to the
interpersonal purposes of conversation.
The most surprising nding to emerge from these previous grammatical
studies is the extent to which classroom teaching is similar to conversation,
rather than combining the linguistic characteristics of speech and informa-
tional writing. (The Multi-Dimensional analysis presented in Biber et al.
(2002) shows these same patterns with respect to a much wider range of
linguistic features.) This pattern of use contradicts our prior expectations.
Given that classroom teaching is clearly pre-planned and has a primary
informational focus, we expected that it would use grammatical features in a
similar way to written registers like textbooks. Instead, we found that
classroom teaching was most strongly inuenced by its òral' situational
characteristicsthe real-time production circumstances, the interactions
among participants, and a focus on the speakers' personal concerns.
Given the importance of the speech/writing distinction for grammatical
Figure 1: Distribution of nouns, verbs, and personal pronouns across

registers
features, we expected that this dierence would be equally important for the
use of lexical features. Thus, we predicted that classroom teaching and
conversation would be sharply distinguished from textbooks and academic
prose in the use of lexical bundles. However, as the following sections show,
these registers have surprisingly dierent patterns of use for lexical bundles.
In particular, classroom teaching incorporates both òral' lexical bundles and
`literate' lexical bundles, in contrast to its reliance on stereotypically spoken
patterns of use for grammatical features.
4 LEXICAL BUNDLES IN UNIVERSITY CLASSROOM

TEACHING AND TEXTBOOKS
As can be seen in Figure 2, there are 43 lexical bundles in conversation; 84 in
classroom teaching; 27 in textbooks; and 19 in academic prose. Overall, then,
Figure 2 shows that the two spoken registers use a much greater range of
dierent lexical bundles than the written registers. However, the most
surprising nding here is for classroom teaching: rather than being
intermediate between conversation and the written registers, classroom
teaching far exceeds conversation in the number of dierent lexical bundles.
Figure 3 shows that lexical bundles occur most frequently in classroom
teaching as well, although they are almost as frequent in conversation. Taken
together, Figures 2 and 3 show that classroom teaching uses a large set of
dierent lexical bundles, while conversation relies on the extremely frequent
use of a smaller set of bundles. For example, conversation uses 44 high-
frequency bundles (occurring more than 60 times per million words). In
contrast, classroom teaching has only 28 bundles that occur more than 60
times per million words.
We can begin to explain these patterns by considering the structural
Figure 2: Number of dierent leixal bundles across registers

Figure 3: Overall frequency of lexical bundles across registers
characteristics of lexical bundles found in each register. We noted in section 2

that lexical bundles have strong grammatical correlates, even though they are
not usually complete structural units. In the present study, we identied three
major structural types of lexical bundle, shown in Table 2.
Type 1 bundles incorporate verb phrase fragments. For example, Types 1a
and 1b begin with a subject pronoun followed by a verb phrase (e.g. it's going
to be). Type 1c begins with a discourse marker followed by a verb phrase (e.g.
you know this is). Types 1d and 1e simply begin with a verb phrase (e.g. is going
to be), while Types 1f and 1g are question fragments.
Type 2 bundles incorporate dependent clause fragments in addition to
simple verb phrase fragments. Type 2a bundles (e.g. I want you to) begin with a
main clause followed by a complementizer (e.g. to, if) or a WH-word
introducing a dependent clause. Other Type 2 bundles are dependent clause
fragments beginning with a complementizer or subordinator (e.g. to be able to,
if we look at, what I want to).
In contrast to Types 1 and 2, which have clausal components, Type 3
bundles are phrasal. Types 3a3c consist of noun phrase components, usually
ending with the start of a postmodier (e.g. the end of the, those of you who).
Type 3d consists of prepositional phrase components with embedded
modiers (e.g. of the things that), while Type 3e incorporates comparative
expressions.
Figure 4 shows the distribution of these structural types across registers. Our
earlier research on conversation and academic prose showed that the
grammatical correlates of lexical bundles in conversation are strikingly
dierent from those in academic prose. In conversation, almost 90 per cent
of all common lexical bundles incorporate verb phrases. In fact, approximately
50 per cent of these lexical bundles begin with a personal pronoun + verb
phrase (such as I was going to, I thought that was). An additional 19 per cent of
Table 2: Structural types of lexical bundles

1. Lexical bundles that incorporate verb phrase fragments
1a. (connector +) 1st/2nd person pronoun + VP fragment

Example bundles: you don't have to, I'm not going to, well I don't know
1b. (connector +) 3rd person pronoun + VP fragment
Example bundles: it's going to be, that's one of the, and this is a
1c. Discourse marker + VP fragment
Example bundles: I mean you know, you know it was, I mean I don't
1d. Verb phrase (with non-passive verb):
Example bundles: is going to be, is one of the, have a lot of, take a look at
1e. Verb phrase with passive verb:
Example bundles: is based on the, can be used to, shown in gure N
1f. yesno question fragments:
Example bundles: are you going to, do you want to, does that make sense
1g. WH-question fragments:
Example bundles: what do you think, how many of you, what does that mean
2. Lexical bundles that incorporate dependent clause fragments
2a. 1st/2nd person pronoun + dependent clause fragment

Example bundles: I want you to, I don't know if, I don't know why, you might want
to
2b. WH-clause fragments:
Example bundles: what I want to, what's going to happen, when we get to
2c. If-clause fragments:
Example bundles: if you want to, if you have a, if we look at
2d. (verb/adjective+) to-clause fragment
Example bundles: to be able to, to come up with, want to do is
2e. That-clause fragments:
Example bundles: that there is a, that I want to, that this is a
3. Lexical bundles that incorporate noun phrase and prepositional phrase fragments
3a. (connector +) Noun phrase with of-phrase fragment:

Example bundles: one of the things, the end of the, a little bit of
3b. Noun phrase with other post-modier fragment:
Example bundles: a little bit about, those of you who, the way in which
3c. Other noun phrase expressions:
Example bundles: a little bit more, or something like that, and stu like that
3d. Prepositional phrase expressions:
Example bundles: of the things that, at the end of, at the same time
3e. Comparative expressions:
Example bundles: as far as the, greater than or equal, as well as the
the bundles consist of an extended verb phrase fragment (e.g. have a look at),
while another 17 per cent of the bundles are question fragments (e.g. can I
have a). In contrast, the lexical bundles in academic prose are phrasal rather
than clausal. Almost 70 per cent of the common bundles in academic prose
consist of noun phrase expressions (e.g. the nature of the) or a sequence that
bridges across two prepositional phrases (e.g. as a result of).
Classroom teaching uses about twice as many dierent lexical bundles as
conversation, and about four times as many as textbooks. The distribution
across structural patterns shown in Figure 4 helps explain this extremely
frequent pattern of use. In marked contrast to the general patterns of use for
grammatical features (described in section 3 above), classroom teaching
relies on the lexical bundles associated with both spoken and written
registers. Similar to conversation, classroom teaching makes dense use of
lexical bundles that represent declarative and interrogative clause fragments.
At the same time, classroom teaching is similar to academic prose and
textbooks in making dense use of noun phrase and prepositional phrase
lexical bundles. Thus, the extremely high density of lexical bundles in
classroom teaching exists because this register relies heavily on both òral'
and `literate' bundles.
In addition, classroom teaching has a large inventory of lexical bundles
associated with dependent clause fragments, especially conditional adverbial
clauses and complement clauses (26 dierent bundles in classroom teaching,
versus 16 in conversation, and only two in academic prose, and two in
textbooks). This pattern is surprising, given previous claims that dependent
clauses are more typical of written prose than speech (e.g. O'Donnell 1974;
Kroll 1977; Chafe 1982; Akinnaso 1982; Gumperz et al. 1984). However, it
turns out that these lexical bundles are more common in both classroom
teaching and in conversation than in the written registers (similar to the
Figure 4: Distribution of lexical bundles across structural types

general grammatical patterns of use for verb+complement clause construc-

tions; see Biber et al. 1999, ch. 9). We return to a discussion of these bundles
in section 5.5 below.
Textbooks and academic prose are at the opposite extreme from classroom
teaching on Figures 2 and 3.6 It is surprising that textbook authors do not
incorporate more lexical bundles in their writing, given the heavy reliance on
bundles in classroom teaching. Reasons for this absence might be that
textbook authors tend to use fuller expressions, preferring full clauses rather
than phrasal lexical bundles, as well as the fact that textbook authors are free
of the real-time production constraints of face-to-face teaching and therefore
make more diverse language choices. At the present stage of our analysis, we
are able only to note this distributional dierence between textbooks and
classroom teaching. A much fuller analysis of textbook discourse structures is
required to interpret the relative absence of lexical bundles in that register.
Given that the structural correlates of lexical bundles in these four registers
are so dramatically dierent, it will come as no surprise that their typical
discourse functions dier as well. We turn to a discussion of those functions in
the following section.
5 A PRELIMINARY FUNCTIONAL TAXONOMY OF THE

LEXICAL BUNDLES IN UNIVERSITY CLASSROOM TEACHING
AND TEXTBOOKS
In the present section, we turn to the discourse functions of lexical bundles,
revising and extending our earlier functional taxonomy of lexical bundle
types developed for conversation and academic prose (Biber et al. 2003). For
this analysis, we examined concordance listings to analyse the functions of
common bundles in their extended discourse contexts, focusing especially on
the lexical bundles used in classroom teaching and textbooks. We have two
immediate goals in doing this analysis: (1) to develop a taxonomy that
captures the major discourse functions served by lexical bundles, and (2) to
describe the extent to which each register uses lexical bundles for each of
these functions.
Our approach to developing the functional taxonomy is primarily inductive.
That is, we group together bundles that serve similar functions, based on the
typical meanings and uses of each bundle. We used concordance listings to
examine the use of each bundle in its discourse contexts. Once the bundles
were assigned to groups, we attempted to determine the discourse functions
associated with each of the groups. Of course this last step was inuenced by
previous theoretical studies on the discourse functions of linguistic features
(e.g. Hymes 1974: 22; Halliday 1978; Brown and Fraser 1979; Biber 1988:
33, 1995).
In some cases, a single bundle has multiple functions even in a single
occurrence. For example, bundles like take a look at and let's have a look
function as both directives and topic introducers. In other cases, a single
bundle serves dierent functions depending on the context. For example,

bundles like the beginning of the and at the end of can function as a time
reference, place reference, or text deictic reference. In general, however, most
bundles have a primary function. We have examined potentially multi-
functional bundles in concordance listings and classied them according to
their most common use. (Several of these cases are discussed in 5.15.3
below.)
We distinguish among three primary functions served by lexical bundles in
these registers: (1) stance expressions, (2) discourse organizers, and (3)
referential expressions. Stance bundles express attitudes or assessments of
certainty that frame some other proposition. Discourse organizers reect
relationships between prior and coming discourse. Referential bundles make
direct reference to physical or abstract entities, or to the textual context itself,
either to identify the entity or to single out some particular attribute of the
entity as especially important. Each of these categories has several sub-
categories associated with more specic functions and meanings. Table 3 lists
the bundles grouped into each of these functional categories, showing the
distribution of each bundle across registers. The following subsections describe
each of these categories in detail.
Table 3: Functional classication of common lexical bundles across registers

Classroom Textbooks Conversation Academic
teaching prose
I. STANCE EXPRESSIONS
A. Epistemic stance
Personal:
I don't know if ** **
I don't know what ** ***
I don't know how ** **
I don't know I **
and I don't know ** .
I think it was ** **
and I think that **
you know what I ** **
I don't think so * **
I thought it was * **
well I don't know **
I don't know whether **
I don't know why **
oh I don't know **
Impersonal:
are more likely to ** .
the fact that the . ** **

teaching prose
B. Attitudinal/modality stance
B1) desire
Personal:
if you want to *** **
I don't want to ** ***
do you want to * ***
you want to go **
do you want a **
what do you want **
B2) obligation/directive
Personal:
I want you to ***
you don't have to ** **
you don't want to ** **
you have to be ** .
you have to do ** *
you look at the **
you might want to **
you need to know **
and you have to ** .
going to have to ** **
you want me to **
do you want me **
Impersonal:
it is important to ** *
it is necessary to **
B3) Intention/prediction
Personal:
I'm not going to ** **
we're going to do **
we're going to have ** *
and we're going to **
I was going to * ***
what we're going to **
are we going to * **
are you going to * ***
Impersonal:
it's going to be ** **
is going to be ***
are going to be **
going to be a ** **
going to be the **
Table 3: cont.
teaching prose
not going to be ** *
going to have a * **
B4) ability
Personal:
to be able to *** . . *
to come up with **
Impersonal:
can be used to ** *
it is possible to * **
II. DISCOURSE ORGANIZERS

A. Topic introduction/focus
what do you think ** **
if you look at **
take a look at **
if you have a **
if we look at **
going to talk about **
to look at the **
to go ahead and **
I want to do ** *
what I want to ** .
want to do is **
want to talk about **
you know if you **
a little bit about **
I would like to * **
in this chapter we **
I/I'll tell you what **
have a look at **
let's have a look **
do you know what **
B. Topic elaboration/clarication
has to do with ** .
to do with the ** . . .
I mean you know ** .
you know I mean ** .
nothing to do with * . **
on the other hand ** ** ***
as well as the . ** *
know what I mean **

teaching prose
was going to say **

what do you mean **
III. REFERENTIAL EXPRESSIONS

A. Identication/focus
that's one of the **
and this is a **
and this is the **
is one of the ** ** *
was one of the ** .
one of the things ***
and one of the **
one of the most * ** **
those of you who **
of the things that ***
B. Imprecision
or something like that ** **
and stu like that **
and things like that ** *
C. Specication of attributes
C1) Quantity specication
there's a lot of ** *
have a lot of **
and a lot of **
a lot of people **
a lot of the **
how many of you **
in a lot of **
the rest of the ** ** * *
a little bit of **
a little bit more **
a lot of times **
than or equal to ** .
greater than or equal **
per cent of the **
C2) Tangible framing attributes
the size of the . ** *
in the form of . ** **
C3) Intangible framing attributes
the nature of the . ** **
in the case of ** ** ***
in terms of the ** ** *
Table 3: cont.
teaching prose
as a result of * ** **
on the basis of . ** **
in the absence of * **
the way in which * * **
the extent to which * **
in the presence of **
D. Time/place/text reference
D1) Place reference
the United States and * **
in the United States ** *** *
of the United States * **
D2) Time reference
at the same time ** ** * **
at the time of . * **
D3) Text deixis
shown in gure N ** *
as shown in gure ** *
D4) Multi-functional reference
the end of the ** ** ** **
the beginning of the * ** *
the top of the * ** . .
at the end of ** *** ** **
in the middle of ** * *
IV. SPECIAL CONVERSATIONAL

FUNCTIONS
A. Politeness
thank you very much **
B. Simple inquiry
what are you doing **
C. Reporting
I said to him/her **
Key to symbols:
. = 1019 per million words
* 2039 per million words
** = 4099 per million words
*** = over 100 per million words
5.1 Stance bundles

Stance bundles provide a frame for the interpretation of the following
proposition, conveying two major kinds of meaning: epistemic and attitude/
modality. Epistemic stance bundles comment on the knowledge status of the
information in the following proposition: certain, uncertain, or probable/
possible (e.g. I don't know if, I don't think so). Attitudinal/Modality stance
bundles express speaker attitudes towards the actions or events described in
the following proposition (e.g. I want you to, I'm not going to). Stance bundles
can be personal or impersonal. Personal stance bundles are overtly attributed
to the speaker/writer, as in the examples given above. Impersonal stance
bundles express similar meanings without being attributed directly to the
speaker/writer (e.g. it is possible to, can be used to).
5.1.1 Epistemic stance bundles

Personal epistemic bundles: Most epistemic stance bundles are personal.
Although epistemic stance bundles can express certainty or uncertainty,
most of these bundles express only uncertainty, as in:
That's, kind of hard to tell, but again, the important thing is be
resourceful when you do these. I don't know what, I don't know what the
voltage is here, so, but, the real point is it's irrelevant. (classroom
teaching)
Expressions with I (don't) think express possibility but a lack of certainty. Thus
compare the two bundles in the following:
I don't know if it will mean revolution in the same sense of the word, I
don't think so because I think there are other political factors involved.
(classroom teaching)
Several lexical bundles in classroom teaching combine epistemic stance with

other functions. For example, bundles with I think/thought it serve a dual
function of referential identication (see III.A in Table 3, discussed in 5.3
below) combined with an uncertain epistemic stance; for example:
The Wall Street Journal last week or I think it was the Wall Street
Journal had something about NASA and this same problem. (classroom
teaching)
Imprecision bundles like and stu like that, discussed in 5.3 below, also serve
an epistemic function combined with referential identication. The bundle
what do you think functions to introduce a new topic but also expresses an
epistemic stance (see 5.2 below).
Impersonal epistemic bundles: In contrast to the personal epistemic bundles,
the impersonal epistemic stance bundles express degrees of certainty rather
than uncertainty.
Boys are more likely to be hyperactive, disruptive, and aggressive in class.

(textbook)
Yet there was irony in the fact that the Russian Revolution, one of the
most important Western revolutions, proclaimed itself to be Marxist in
aims and character but happened in violation of Marxist historical logic.
(textboook)
5.1.2 Attitudinal/modality stance bundles

Attitudinal/modality stance bundles are also usually personal, expressing
speaker attitudes towards the actions or events described in the following
proposition. Four major subcategories are distinguished here: desire, obliga-
tion/directive, intention/prediction, and ability.
Desire bundles: Desire bundles include only personal expressions of stance,
which frame self-motivated wishes and desires or inquire about another
participant's desires:
So I may not want to see her face to face because I don't want to deliver
bad news to her. (classroom teaching)
Several lexical bundles that express personal desire in classroom teaching are
also used to initiate new topics, including what I want to do and I would like to;
these are discussed in section 5.2 below.
Obligation/directive bundles: The second subcategory of attitudinal/modality
stance bundles expresses obligations or directives. Most of these bundles are
personal stance expressions, but they dier from other personal bundles in
that they have a second person pronoun (you) rather than rst person
pronoun as subject. However, they are still clearly understood as personal
expressions of stance, directing the listener to carry out actions that the
speaker wants to have completed. For example:
Now you need to know how to read these. (classroom teaching)
All you have to do is work on it. (classroom teaching)
In some cases, these bundles include verbs of desire with a rst person
pronoun, directly conveying the speaker's desire that the addressee carry out
some action, and thus functioning as directives:
I want you to take out a piece of paper and jot some notes down in these
four areas. (classroom teaching)
In other cases, the directive force of some of these bundles can be very
indirect, as in:
You might want to look at a couple of examples just to remind yourself of
how these look. (classroom teaching)
Several bundles can t either the desire category or obligation/directive,
depending on the context of use. For example, the following would be a desire
bundle if it truly is a question about the participant's preference, but it is more

likely to function as a directive when said by a teacher to a student:
So, do you want to be next Erin? (classroom teaching)
Some directive bundles are used for topic introduction (e.g. take a look at);
these are discussed in 5.2 below.
A few obligation/directive stance bundles are impersonal, with no personal
pronoun at all, even though they still clearly direct the reader to carry out
some action:
It is important to note that Derrida does not assert the possibility of
thinking outside such terms. (textbook)
Intention/prediction bundles: The third subcategory of attitudinal/modality
stance bundles is intention/prediction. Many of these bundles are overtly
personal, expressing the speaker's own intention to perform some future
action. In most cases, these are expressions of joint action, used to announce
the proposed plan of a class session, as in:
But, right now what we're going to take a look at are ones that are
produced that are positive and benecial. (classroom teaching)
Other bundles in this category are impersonal, expressing predictions of future
events that do not entail the volition of the speaker. These bundles are usually
used when explaining a logical or mathematical process that involves several
steps, as in:
And so if you require a, twenty percent return on investment, this net
present value is going to be zero. (classroom teaching)
Ability bundles: Only four attitudinal/modality stance bundles express ability.
In their form, all four bundles are impersonal, although these bundles
sometimes occur together with directive bundles to identify skills and tasks
that students should accomplish:
I want you to be able to name and dene those four curriculum category
[sic]. (classroom teaching)
So encoding's always harder than decoding. Cos you have to come up
with the word, you have to spell it, you have to use it correctly.
5.2 Discourse organizing bundles

Discourse organizing bundles serve two major functions: topic introduction/
focus and topic elaboration/clarication.
5.2.1 Topic introduction/focus bundles

Topic introduction bundles in classroom teaching provide overt signals to the
student that a new topic is being introduced. Many of these are expressions of
intention or desire (see 5.1 above), but they have the more specialized
function of announcing the instructor's intention to begin a new topic:7
But, before I do that, I want to talk about Plato. (classroom teaching)
What I want to do is quickly run through the exercise that we're going to
do. OK just so you see what it does. (classroom teaching)
As the preceding example shows, sometimes two four-word bundles occur

together, in eect creating a longer ve-word or six-word bundle.
The following example illustrates the use of these longer bundles for
procedural instructions, identifying the major steps in the procedure:
OK? Next thing I want to do iswhat I want to do is I want to change the
back color [. . .] OK? First thing I want to do is let's set up some colors of
the text boxes to start with [. . .] OK? First thing I want to do is let's
make the rst text box. (classroom teaching)
Examples like this provide strong evidence to support the claim that lexical
bundles are stored as unanalysed units in the mental lexicon. These topic
introducing bundles often result in `syntactic blends' (see Biber et al. 1999:
10646): the bundle signals that a new topic (or step in the procedure) is
coming up, followed by the statement of the topic, but the two parts are
not well-formed syntactically. For example, the bundle used repeatedly in
the excerpt aboveI want to do isrequires an obligatory subject
predicative to make it syntactically complete. However, what we nd
instead is that the speaker starts over with a new syntactic construction (a
let's imperative in the last two occurrences). The bundle here serves the
discourse function of identifying a new topic, and it therefore functions as if
it were syntactically complete; the following constituent begins a new main
clause, rather than completing the subject predicative structure required by
the bundle.
Topic introducing bundles can occur with both rst and second person
pronouns. The rst person plural pronoun we as subject seems to invite
student participation, although the `we' often refers to the instructor rather
than a collective enterprise:
Today we are going to talk about testing hypotheses. (classroom
teaching)
Now, we want to talk about getting our sample mean . . . (classroom
teaching)
Topic introducing bundles with second person pronouns invite student

participation, although the instructor is usually intending collective con-
sideration of the topic:
If you look at development and the jobs that are created, it says nothing
rst of all of the type of jobs that are created. (classroom teaching)
The bundle if you look at often has a deictic reference, identifying the props
required for a topic. The bundle directs students' attention to the prop,
indirectly introducing a new topic by reference to it:
If you look at the answers that are given, there's only two answers that
have one big M . . . (classroom teaching)
Finally, topic introducing bundles with WH-question structures provide the
most overt attempts to directly engage students in a new topic:
What do you think the text is trying to tell us when they call our
attention that often conict doesn't appear suddenly? (classroom
teaching)
5.2.2 Topic elaboration/clarication bundles

The second major subcategory of discourse organizing bundles relates to topic
elaboration or clarication. For example,
Well why is the Navajo Depot, Camp Navajo important today? [. . .] It
has to do with the START talks with the Russians, the START Treaty
signed in 1991. (classroom teaching)
The discourse markers you know and I mean are used in sequence as a lexical
bundle, usually when the speaker believes that additional explanation or
clarication is required:
When you come to class next timeand I'm gonna look at grammar
you know I mean I expect you to have things spelled relatively correctly
. . . (classroom teaching)
The bundles as well as the and on the other hand are used for explicit comparison
and contrast. These two discourse organizing bundles are considerably more
common in textbooks than in classroom teaching:
Section 3.5 [. . .] illustrates how the techniques are employed together
as well as the range of resulting execution characteristics that are
presented to an architecture, . . . (textbook)
We know that if the project is in the same line business as the rm's
other projects [. . .] then high stand alone risk translates into high
corporate risk [. . .] On the other hand, if the project is not in the same
line business, then it is possible that the correlation may be low . . .
(textbook)
5.3 Referential bundles

Referential bundles generally identify an entity or single out some particular
attribute of an entity as especially important. We describe four major
sub-categories included under referential bundles: identication/focus,
imprecision indicators, specication of attributes, and time/place/text refer-

ence.
5.3.1 Identication/focus bundles

Identication/focus bundles are common in classroom teaching, focusing on
the noun phrase following the bundle as especially important. For example,
the bundle those of you who identies the subgroup of students who are in
focus:
For those of you who came late I have the, uh, the quiz. (classroom
teaching)
In many cases, identication/focus bundles also have a discourse organizing

function (see 5.2). These bundles are often used after a lengthy explanation to
emphasize or summarize the main point:
Schizophrenia typically uh will mean that uh separation from reality
uh it can mean uh uh you know extreme periods of euphoria and
extreme periods of depression it can mean a lot of thingsand that's one
of the problems of schizophrenia. (classroom teaching)
OK. Uh we create a tri-block for an object of type thread, and there is a

built-in thread object that has a method called sleep, and that method
called sleep takes a parameter which is the number of milliseconds [. . .]
OK? And this is a real simple way, the simplistic way to do animation.
In other cases, identication/focus bundles can be used to introduce a

discussion by stating the main point rst, and then giving the details:
One of the things they stress in parenting is to be consistent and
particularly with parents um some parents are inconsistent between
siblings. Uh fathers are notorious for letting their little darling girls get
away with what they swat the boys about . . . (classroom teaching)
5.3.2 Imprecision bundles

A second major subcategory of referential bundles indicates imprecise
reference. These have two specic functions, either to indicate that a specied
reference is not necessarily exact, or to indicate that there are additional
references of the same type that could be provided:
I think really we now have what about, six weeks left in class or
something like that. (classroom teaching)
There are obviously companies that do uh evaluations and things like

that (classroom teaching)
5.3.3 Bundles specifying attributes

The third subcategory of referential bundles identify specic attributes of the
following head noun. Some of these bundles specify quantities or amounts:
You'd have a lot of power. (classroom teaching)
Does it create a lot of wealth? No. It creates a little bit of wealth.
The bundle a little bit about usually has a more specialized function to
introduce a topic (see 5.2 above), apparently to minimize the expectations
required from students:
So I want to talk a little bit about process control from that point of view.
Other bundles in this category describe the size and form of the following
head noun:
These gures give an idea of the size of the ethnological community in
Russia. (textbook)
They are in the form of half-wheels, with concentric bands of
representations alternating with bands of scrollwork. (textbook)
In contrast, some specifying bundles identify abstract characteristics:
Rather than reading textbooks and solving textbook problems, students
must dene and constantly rene the nature of the problem . . .
(textbook)
These abstract specifying bundles are often used to establish logical relation-
ships in a text:
Fleshy fruits are classied on the basis of the dierentiation of the fruit
wall (pericarp). (textbook)
They are dened in terms of the emotion they elicit. (textbook)
5.3.4 Time/place/text-deixis bundles

Finally, several referential bundles refer to particular places, times, or
locations in the text itself. Three place bundles in our corpus refer to the
United States, reecting one focus of textbooks and classroom teaching in our
corpus:
Children in the United States are not formally employed in farm work, . . .
(textbook)
Text deixis bundles are common only in the written registers, where they
make direct reference to gures contained in the text itself:
As shown in Figure 4.4, the higher the real estate agents scored in terms
of the proactive personality dimension, the more houses they sold . . .
(textbook)
Many of these bundles are multi-functional, referring to a place, time, and/or
text deixis, depending on the particular context:
So you have to record that, since the asset was sold at the end of the year
She's in that.. uh.. oce down there.. at the end of the hall (classroom
teaching)
uh I'm going to start actually with the end of the chapter (classroom
teaching)
5.4 Register variation in the functional exploitation of lexical

bundles
In the preceding sections, we have outlined a taxonomy of the major
discourse functions served by lexical bundles. The taxonomy was developed to
include functions that can potentially be realized in any register. However, as
Table 4 shows, the four registers show dramatic dierences in their reliance on
particular functional types. The examples presented in sections 5.15.3
illustrate the use of these bundles in their characteristic registers.
Figure 5 provides a graphic representation of these data, showing the
following overall patterns:
1 stance bundles are extremely common in both classroom teaching and
conversation;
2 discourse organizing bundles are most common in classroom teaching and
somewhat less common in conversation;
3 referential bundles are extremely common in classroom teaching and
somewhat less common in textbooks and academic prose.
The patterns of use in classroom teaching are especially interesting here, and
they help to explain why lexical bundles are generally so much more
common in this register than any other register.8 Classroom teaching
Table 4: Distribution of common lexical bundles across functions

Conversation Classroom Textbooks Academic
teaching prose
Stance bundles 29 33 4 3
Discourse organizers 10 19 3 1
Referential bundles 3 32 20 15
Figure 5: Distribution of lexical bundles across functional categories
combines functional characteristics of both conversation (using stance and

discourse organizing bundles) and textbooks/academic prose (using referential
bundles). However, classroom teaching actually goes beyond these other
registers, using bundles in all three functional categories to a greater extent
than any other register.
Two major patterns are noteworthy here: rst, classroom teaching
combines the functional and communicative priorities of involved spoken
discourse (shown by the dense use of stance bundles) with the priorities of
informational written discourse (shown by the dense use of referential
bundles). Second, classroom teaching is structured with lexical bundles to a
greater extent than any other register. This is shown most clearly by the large
number of discourse organizing bundles used in classroom teaching. (Many of
the referential bundles in classroom teaching are also used for discourse
functions, such as identication/focus, imprecision, and quantity specica-
tion.) In fact, classroom teaching actually uses the most bundles in each
functional category. This pattern apparently reects the complex commun-
icative demands of this register. Lexical bundles are useful for instructors who
need to organize and structure discourse that is at once informational and
involved, and is produced with real-time production constraints.
5.5 The relationship between structural and functional

categories
Figure 6 shows that there is a very strong relationship between structural type
and discourse function for lexical bundles. In particular, most stance bundles
are composed of dependent clause fragments, while most referential bundles
are composed of noun phrase or prepositional phrase fragments. (The
prediction/intention stance bundles are all composed of VP fragments,
Figure 6: Interaction of structural and functional categories
mostly incorporating the semi-modal be going to.) Discourse organizers are the
only functional category to use all three structural types.
Obviously, these patterns are strongly associated with register: conversa-
tion, at the òral' extreme, uses mostly VP-based and dependent clause lexical
bundles for stance functions; academic prose, at the `literate' extreme, uses
mostly NP/PP-based lexical bundles for referential functions. Although
classroom teaching is generally similar to conversation in its grammatical
characteristics (see section 3 above), it uses NP/PP-based lexical bundles for
referential purposes.
These patterns suggest that there is a direct association between form and
function for lexical bundles. For example, complement clause-based bundles
are used for stance functions, because complement clauses are generally one
of the most productive grammatical devices used to express stance in English
(see Biber et al. 1999, ch. 12). Thus, it makes sense that many of the most
common multi-word sequences with complement clauses would become
xed as stance lexical bundles. Similarly, noun phrases and prepositional
phrases are the primary grammatical devices used for referential functions,
and so it makes sense that the most common of these multi-word sequences
would become xed as referential lexical bundles. Thus, we see a complex
interaction between structural form, discourse function, and the typical
purposes and situational characteristics of registers.
6 CONCLUSION: THE THEORETICAL STATUS OF LEXICAL

BUNDLES
The results of the present analysis suggest that lexical bundles should be
regarded as a basic linguistic construct with important functions for the
construction of discourse. However, with respect to both structure and
function, lexical bundles dier dramatically from other linguistic features

(including the traditional formulaic expressions usually recognized by
discourse analysts).
Given that lexical bundles are dened strictly on the basis of frequency,
with no consideration of structural or functional criteria, they might be
expected to be arbitrary strings of words that have no linguistic status. Instead,
these frequent sequences of words turn out to be readily interpretable in both
structural and functional terms. Although they are not the kinds of
grammatical structures recognized by traditional linguistic theory, most lexical
bundles do have well-dened structural correlates: they usually consist of the
beginning of a clause or phrase plus the rst word of an embedded structure
(e.g. a dependent complement clause or a prepositional phrase). These
sequences of words can be regarded as structural `frames', followed by a `slot'.
The frame functions as a kind of discourse anchor for the `new' information in
the slot, telling the listener/reader how to interpret that information with
respect to stance, discourse organization, or referential status.
The patterns of use for lexical bundles are notably dierent from those
found for traditional grammatical features. The contrast is especially striking
for classroom teaching. With respect to grammatical features, classroom
teaching relies heavily on òral' structures, despite the need for an
informational focus. In our 2002 Multi-Dimensional study, we note that:
On the spoken end, all university registers had scores indicating high
involvement, reecting their frequent use of such features as present
tense verbs, private verbs, rst- and second-person pronouns, and
contractions. The most surprising inclusion in this group was classroom
teaching, which had a notably involved rather than informational
characterization. (Biber et al. 2002: 27)
These ndings indicate that the typical grammatical characteristics of class-

room teaching are determined primarily by the òral' characteristics of the
situation: the real-time production circumstances, and the focus on personal
and interpersonal purposes.
In contrast, in the present study we found that classroom teaching mixed
òral' and `literate' characteristics in the use of lexical bundles, actually going
beyond the expected `targets' in its patterns of use. That is, classroom teaching
shows a more extensive use of stance lexical bundles and discourse organizing
bundles than in conversation, while at the same time it shows a more
extensive use of referential bundles than in academic prose.
These patterns of use indicate that lexical bundles are a fundamentally
dierent kind of linguistic construct from productive grammatical construc-
tions. For example, consider the use of NP/PP-based referential bundles in
contrast to other noun- and prepositional-phrase structures. Classroom
teaching relies heavily on NP/PP-based referential bundles but avoids the
dense use of noun phrase and prepositional phrase structures in general. As a
productive grammatical strategy, the dense use of complex noun phrase
constructions has dramatically increased in informational written registers

over the past 100 years (see Biber and Clark 2002; Biber 2003). However,
these constructions are much less common in spoken registers like
conversation, presumably because they are dicult to produce and
comprehend in real-time situations. In this regard, classroom teaching is
typical of other òral' registers in avoiding the dense use of complex noun
phrase structures.
Given those grammatical patterns, it is surprising that classroom teaching
makes extensive use of NP/PP-based referential bundles. Although the
functional need for referential expressions in classroom teaching is clear,
the reliance on NP/PP-based bundles is unexpected. We interpret this pattern
as evidence that lexical bundles are stored as unanalysed multi-word chunks,
rather than as productive grammatical constructions. The fact that referential
bundles are composed of noun phrase and prepositional phrase fragments
reects their historical origins, but we would argue that these sequences are
now stored and used as single units, without reference to their structural
correlates. As such, these bundles do not present production or comprehen-
sion diculties for speakers and listeners in classroom teaching.
More general evidence for the importance of lexical bundles comes from
their frequencies of use and obvious discourse functions. We have approached
the study of lexical bundles, like all our corpus studies, with the general
hypothesis that high frequency patterns are not accidental, nor are they
explanatory. Rather, corpus-based frequency evidence provides descriptive
facts that require explanation. In the present study, the facts that require
explanation are the existence of common multi-word sequences which do not
represent well-dened linguistic structures. Examination of these multi-word
sequences in textual contexts shows that they are important building blocks of
discourse, associated with basic communicative functions. In general, these
lexical bundles serve as discourse framing devices: they provide a kind of
frame expressing stance, discourse organization, or referential status,
associated with a slot for the expression of new information relative to that
frame.
Obviously, other approaches with dierent goals are important to
complement the present study and increase our understanding of multi-
word units, including psycholinguistic studies and investigations of more
idiomatic and perceptually salient expressions.9 At the same time, this study
illustrates how an exploratory corpus approach facilitates the identication of
language features that would go unrecognized otherwise, but that turn out to
be a fundamentally important part of writers' and speakers' communicative
repertoire.
(Final version received April 2004)
NOTES
1 This study was carried out as part of the picture) occur only 10 times per million
TOEFL 2000 Spoken and Written Academic words.
Language Corpus project, supported by 5 The quantitative ndings for conversation
Educational Testing Service, with the goal and academic prose are taken from Biber et
of providing a basis for test construction and al. 1999; those for classroom teaching and
validation (see Biber et al. 2004). textbooks are taken from Biber et al. (2004).
2 The quantitative analysis of lexical bundles 6 Figure 2 shows that textbooks uses a greater
was undertaken with computer programs range of dierent lexical bundles than
that identied and stored every four-word academic prose. Some of these bundles in
sequence in our corpus. The programs read academic prose occur with high frequencies,
through each text in the corpus, storing resulting in the slightly higher overall
every sequence beginning with the rst frequency of lexical bundles shown in
word of the text and advancing one word Figure 3.
at a time. For example, the rst sentence of 7 In fact, some of these bundles typically have
this note would have the following four- stance functions in conversation, while they
word sequences identied: the quantitative usually serve discourse organizing functions
analysis of, quantitative analysis of lexical, in classroom teaching. The bundle I would
analysis of lexical bundles, of lexical bundles like to is a good example of this type.
was, etc. Each time a sequence was identi- 8 Some dierences in the sets of lexical
ed, it was automatically checked against bundles found across registers might be due
the previously identied sequences, and a to dierences in the corpora analysed:
running frequency count showed how often ranging from c.5-million words for academic
each sequence was repeated. In identifying prose, to 1.2 million words for classroom
lexical bundles, we relied on orthographic teaching, and .75 million words for text-
word units, even though these sometimes books. However, Cortes (2002: 725) found
arbitrarily combine separate words. For ex- that analyses of smaller corpora actually
ample, into, cannot, self-control, and don't are yield more lexical bundles, because some
all regarded as single words in our analysis. bundles have articially high frequencies in
Only uninterrupted sequences of words the smaller corpora that cannot be main-
were treated as lexical bundles. Thus, lexical tained in a larger collection of texts. Thus,
sequences that spanned a turn boundary or a the low number of lexical bundles found in
punctuation mark were excluded. textbooks cannot be attributed to the fact
3 Other sequences of words can be repeated that that sub-corpus is smaller; if anything,
frequently within a single text. In many we would expect to nd an articially
cases, these other sequences do not repres- inated number of bundles based on analysis
ent lexical bundles, because they are not of a smaller sub-corpus.
widely used across multiple texts. These local 9 Schmitt et al. (2004) directly investigate the
repetitions reect the immediate topical psycholinguistic status of lexical bundles
concerns of the discourse. In contrast, lexical through an experimental design. Their
bundles can be regarded as the more general study indicates that not all bundles are
lexical building blocks that are used fre- stored in the mind as formulaic sequences,
quently by many dierent speakers/writers although register factors are not considered
within a register. in the study. One longer-term goal is to
4 However, Simpson and Mendis (2003) docu- extend the methods used to identify lexical
ment the important pragmatic functions of bundles to allow for variations on a pattern.
idioms in classroom teaching. These are We considered the use of longer bundles and
generally rare, and they are often short variations on a lexical pattern in Biber et al.
noun phrases or prepositional phrases. For (1999, ch. 13). In addition, speakers often
example, the most common idioms identi- use a series of expressions that are related in
ed in this study (bottom line and the big form and function, although they are not
strictly recurrent lexical sequences. For ex- complexity. and I'm gonna show
ample, the following text sample illustrates you some slides of this in just a
variations on a bundle that incorporates the moment but basically we can look
elements 1st/2nd person pronoun + look at: at animals from very simple ani-
mals to complex animals . . .
If we look at the fossils we nd
some very simple animals and if
you look at more and more recent In our future research, we hope to interpret
fo- fossils we nd more and more the use of lexical bundles relative to this
complex animals and what I wider range of lexical expressions. This type
would like you to understand is of study is much more feasible when
what does this mean to increase in focusing on a single lexical bundle in a few
complexity so we'll look at what we selected texts. The problem comes in trying
call evolutionary trends or to identify the full range of lexical bundles
increases [writing on board] in across a large corpus of texts.
REFERENCES
Aijmer, K. 1996. Conversational routines in bundles in conversation and academic
English: Convention and creativity. London: prose' in H. Hasselgard and S. Oksefjell
Longman. (eds): Out of Corpora: Studies in honor of Stig
Akinnaso, F. 1982. Òn the differences Johansson. Amsterdam: Rodopi, pp. 1819.
between spoken and written language.' Biber, D., S. Conrad, and V. Cortes. 2003.
Language and Speech 25: 97125. `Lexical bundles in speech and writing: An
Altenberg, B. 1998. Òn the phraseology of initial taxonomy' in A. Wilson, P. Rayson,
spoken English: The evidence of recurrent and T. McEnery (eds): Corpus Linguistics by
word-combinations' in A. Cowie (ed.): the Lune. Frankfurt/Main: Peter Lang, pp. 71
Phraseology: Theory, analysis and applications. 93.
Oxford: Oxford University Press, pp. 101 Biber, D., S. Johansson, G. Leech, S. Conrad,
22. and E. Finegan. 1999. The Longman Gram-
Altenberg, B. and Eeg-Olofsson. 1990. `Phras- mar of Spoken and Written English. London:
eology in spoken English' in J. Aarts and Longman.
W. Meijs (eds): Theory and Practice in Corpus Biber, D., S. Conrad, R. Reppen, P. Byrd, and
Linguistics. Amsterdam: Rodopi, pp. 126. M. Helt. 2002. `Speaking and writing in the
Biber, D. 1988. Variation across Speech and university: A multi-dimensional compari-
Writing. Cambridge: Cambridge University son.' TESOL Quarterly, 36, 948.
Press. Biber, D., S. Conrad, R. Reppen, P. Byrd,
Biber, D. 1995. Dimensions of Register Variation: M. Helt, V. Clark,V. Cortes, E. Csomay, and
A cross-linguistic comparison. Cambridge: Cam- A. Urzua. 2004. Representing Language Use in
bridge University Press. the University: Analysis of the TOEFL 2000
Biber, D. 2003. `Compressed noun phrase Spoken and Written Academic Language
structures in newspaper discourse: The com- Corpus. TOEFL Monograph Series. Princeton,
peting demands of popularization vs. eco- NJ: Educational Testing Service.
nomy' in J. Aitchison and D. Lewis (eds): Brown, P. and C. Fraser. 1979. `Speech as a
New Media Discourse. London: Routledge, marker of situation' in K. R. Scherer and
pp. 16981. H. Giles (eds): Social Markers in Speech. Cam-
Biber, D. and V. Clark. 2002. `Historical shifts bridge: Cambridge University Press, pp. 33
in modification patterns with complex noun 62.
phrase structures: How long can you go Butler, C. S. 1997. `Repeated word combina-
without a verb?' in T. Fanego, M. J. Lopez- tions in spoken and written text: Some
Couso, and J. Perez-Guerra (eds): English implications for Functional Grammar' in
Historical Syntax and Morphology. Amsterdam: C. S. Butler, J. H. Connolly, R. A. Gatward,
John Benjamins, pp. 4366. and R. M. Vismans (eds): A Fund of Ideas:
Biber, D. and S. Conrad. 1999. `Lexical Recent developments in Functional Grammar.
Amsterdam: IFOTT, University of Amster- 1996. Collins COBUILD Grammar Patterns 1:

dam, pp. 6077. Verbs. London: HarperCollins.
Butler, C. S. 1998. `Collocational frameworks Francis, G., S. Hunston, and E. Manning.
in Spanish.' International Journal of Corpus 1998. Collins COBUILD Grammar Patterns 2:
Linguistics, 3, 132. Nouns and Adjectives. London: HarperCollins.
Chafe, W. L. 1982. Ìntegration and involve- Francis, G., E. Manning, and S. Hunston.
ment in speaking, writing, and oral liter- 1997. Verbs: Patterns and Practice. London:
ature' in D. Tannen (ed.): Spoken and Written HarperCollins.
Language: Exploring orality and literacy. Nor- Gledhill, C. 2000. `The discourse function of
wood, NJ: Ablex, pp. 3554. collocation in research article introductions.'
Chaudron, C. and J. Richards. 1986. `The English for Specific Purposes, 19, 11535.
effect of discourse markers on the compre- Grabe, W. and R. B. Kaplan. 1996. Theory and
hension of lectures.' Applied Linguistics, 7, Practice of Writing. London: Longman.
11327. Grabe, W. and R. B. Kaplan. 1997. Òn the
Chih-Hua, K. 1999. `The use of personal writing of science and the science of writing:
pronouns: Role relationships in scientific Hedging in scientific text and elsewhere' in
journal articles.' English for Specific Purposes, R. Markkanen and H. Schroeder (eds):
18, 12138. Hedging in Discourse. Berlin: de Gruyter,
Conrad, S. and D. Biber. In press. `The pp. 15167.
frequency and use of lexical bundles in Granger, S. 1998. `Prefabricated patterns in
conversation and academic prose, in advanced EFL writing: collocations and
W. Teubert and M. Mahlberg (eds), `The formulae' in A. Cowie (ed.): Phraseology.
corpus approach to lexicography, Thema- Oxford: Oxford University Press, pp. 145
tischer Teil von Lexicographica'. Internatio- 60.
nales Jahrbuch fur Lexicographie 20. Gumperz, J. J., H. Kalman, and M. C.
Cortes, V. 2002. Lexical bundles in academic O'Conner. 1984. `Cohesion in spoken and
writing in history and biology. Unpublished written discourse' in D. Tannen (ed.):
doctoral dissertation, Northern Arizona Uni- Coherence in Spoken and Written Discourse.
versity. Norwood, NJ: Ablex, pp. 320.
Cortes, V. To appear. `Lexical bundles in Hakuta, K. 1974. `Prefabricated patterns and
published and student disciplinary writing: the emergence of structure in second lan-
Examples from history and biology.' English guage acquisition.' Language Learning, 24,
for Specific Purposes. 28797.
Crompton, P. 1997. `Hedging in academic Halliday, M. A. K. 1978. Language as Social
writing: Some theoretical problems.' English Semiotic: The social interpretation of language
for Specific Purposes, 16, 27187. and meaning. London: Edward Arnold.
DeCarrico, J. and J. Nattinger. 1988. `Lexical Halliday, M. A. K. 1988. Òn the language of
phrases for the comprehension of academic physical science' in M. Ghadessy (ed.):
lectures.' English for Specific Purposes, 7, 91 Registers of Written English. London: Pinter,
102. pp. 16278.
deCock, S. 1998. À recurrent word combina- Halliday, M. A. K. and J. R. Martin. 1993.
tion approach to the study of formulae in the Writing Science: Literacy and discursive power.
speech of native and non-native speakers of Pittsburgh: University of Pittsburgh Press.
English.' International Journal of Corpus Lin- Howarth, P. 1996. Phraseology in English Aca-
guistics, 3, 5980. demic Writing. Tubingen: Max Niemeyer
Ellis, N. 1996. `Sequencing in SLA: Phonolo- Verlag.
gical memory, chunking, and points of Howarth, P. 1998a. `Phraseology and second
order.' Studies in Second Language Acquisition, language proficiency.' Applied Linguistics, 19,
19, 91126. 2444.
Flowerdew, J. and S. Tauroza. 1995. `The Howarth, P. 1998b.'The phraseology of learn-
effect of discourse markers of second lan- ers' academic writing' in A. Cowie (ed.):
guage lecture comprehension.' Studies in Phraseology. Oxford: Claredon Press, pp. 161
Second Language Acquisition, 17, 43558. 86.
Francis, G., S. Hunston, and E. Manning. Hudson, J. 1998. Perspectives on Fixedness:
Applied and theoretical. Lund: Lund University Using corpora for English language research
Press. and teaching. Amsterdam: Benjamins.
Hunston, S. and G. Francis. 1998. `Verbs Partington, A. and J. Morley. 2004. `From
observed: a corpus-driven pedagogic gram- frequency to ideology: Investigating word
mar.' Applied Linguistics, 19, 4572. and cluster/bundle frequency in political
Hunston, S. and G. Francis. 1999. Pattern debate' in B. Lewandowska-Tomaszczyk
Grammar: A Corpus-driven approach to the (ed.): Practical Applications in Language and
lexical grammar of English. Amsterdam: Ben- ComputersPALC 2003. Frankfurt a. Main:
jamins. Peter Lang, pp. 17992.
Hyland, K. 1994. `Hedging in academic writing Pawley, A. and H. Syder. 1983. `Two puzzles
and EAP textbooks.' English for Specific for linguistic theory: Native-like selection
Purposes, 13, 23956. and native-like fluency' in J. Richards and
Hyland, K. 1996a. `Talking to the academy: R. Schmidt (eds): Language and Commun-
Forms of hedging in science research art- ication. London: Longman, pp. 191226.
icles.' Written Communication, 13, 25181. Redeker, G. 1991. `Review article: Linguistic
Hyland, K. 1996b. `Writing without convic- markers of discourse structure.' Linguistics,
tion? Hedging in science research articles.' 29, 113972.
Applied Linguistics, 17, 43354. Renouf, A. and J. Sinclair. 1991. `Colloca-
Hymes, D. 1974. Foundations in Sociolinguistics. tional frameworks in English' in K. Aijmer
Philadelphia: University of Pennsylvania and B. Altenberg (eds): English Corpus Lin-
Press. guistics: Studies in honour of Jan Svartvik.
Kjellmer, G. 1991. À mint of phrases' in London: Longman, pp. 12843.
K. Aijmer and B. Altenberg (eds): English Salem, A. 1987. Pratique des Segments Repetes.
Corpus Linguistics: Studies in honour of Jan Paris: Institut National de la Langue Fran-
Svartvik. London: Longman, pp. 11127. caise.
Kroll, B. 1977. `Ways communicators encode Schmitt, N. (ed.). 2004. Formulaic Sequences.
propositions in spoken and written English: Amsterdam: John Benjamins.
A look at subordination and coordination' in Schmitt, N., S. Grandage and S. Adolphs.
E. O. Keenan and T. Bennett (eds): Discourse 2004. Àre corpus-derived recurrent clusters
across Time and Space. Los Angeles: University psycholinguistically valid?' In N. Schmitt
of Southern California, pp. 69108. (ed.), Formulaic Sequences. Amsterdam: John
Lewis, M. 1993. The Lexical Approach: The state Benjamins. pp. 12751.
of ELT and a way forward. Hove: LTP. Simpson, R. and D. Mendis. 2003. À corpus-
Marco, M. J. L. 2000. `Collocational frame- based study of idioms in academic speech.'
works in medical research papers: a genre- TESOL Quarterly, 37, 41941.
based study.' English for Specific Purposes, 19, Sinclair, J. 1991. Corpus, Concordance, Colloca-
6386. tion. Oxford: Oxford University Press.
Moon, R. 1998a. Fixed Expressions and Idioms in Strodt-Lopez, B. 1991. `Tying it all in: Asides
English: A corpus-based approach. Oxford: in university lectures.' Applied Linguistics, 12,
Clarendon. 11740.
Moon, R. 1998b. `Frequencies and forms of Swales, J. M. 1990. Genre Analysis: English in
phrasal lexemes in English' in A. Cowie academic and research settings. Cambridge:
(ed.): Phraseology. Oxford: Oxford University Cambridge University Press.
Press, pp. 79100. Swales, J. M., U. K. Ahmad, Y. Y. Chang,
Myers, G. 1989. `The pragmatics of politeness D. Chavez, D. F. Dressen, and R. Seymour.
in scientific articles.' Applied Linguistics, 10, 1998. `Consider this: The role of imperatives
135. in scholarly writing.' Applied Linguistics, 19,
Nattinger, J. and J. DeCarrico. 1992. Lexical 97121.
Phrases and Language Teaching. Oxford: Thompson, G. and Y. Ye. 1991. Èvaluation in
Oxford University Press. the reporting verbs used in academic papers.'
O'Donnell, R. C. 1974. `Syntactic differences Applied Linguistics, 12, 36582.
between speech and writing.' American Weinert, R. 1995. `The role of formulaic
Speech 49, 10210. language in second language acquisition: A
Partington, A. 1998. Patterns and Meanings: review.' Applied Linguistics, 16, 180205.
Williams, I. 1996. À contextual study of Wray, A. 2002. Formulaic Language and the
lexical verbs in two types of medical research Lexicon. Cambridge: Cambridge University
report: Clinical and Experimental.' English Press.
for Specific Purposes, 15, 17597. Yorio, C. 1980. `Conventionalized language
Wray, A. and M. Perkins. 2000. `The functions forms and the development of commun-
of formulaic language: an integrated model.' icative competence.' TESOL Quarterly, 14,
Language and Communication, 20, 128. 43342.

If You Look at ... Lexical Bundles in University Teaching and Textbooks PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

If You Look at ... Lexical Bundles in University Teaching and Textbooks PDF

Uploaded by

Copyright:

Available Formats

Applied Linguistics 25/3: 371405 # Oxford University Press 2004

If you look at . . .: Lexical Bundles in

This paper investigates the use of multi-word sequences in two important

use of multi-word prefabricated expressions (e.g. in a nutshell, if you see what I

we adopt a frequency perspective`lexical bundles'to investigate the

1.1 Previous research on lexical bundles

1.2 Overview of the present study

Table 1: Composition of the sub-corpus used in the analysis

Classroom teaching 176 1,248,800

Source: The TOEFL 2000 Spoken and Writing Academic LanguageT2K-SWAL

Identication of lexical bundles

3 BACKGROUND: GRAMMATICAL DIFFERENCES AMONG

forming a cline from the more conversational registers to the more

Figure 1: Distribution of nouns, verbs, and personal pronouns across

4 LEXICAL BUNDLES IN UNIVERSITY CLASSROOM

Figure 2: Number of dierent leixal bundles across registers

Figure 3: Overall frequency of lexical bundles across registers

characteristics of lexical bundles found in each register. We noted in section 2

Table 2: Structural types of lexical bundles

1a. (connector +) 1st/2nd person pronoun + VP fragment

2. Lexical bundles that incorporate dependent clause fragments

2a. 1st/2nd person pronoun + dependent clause fragment

3a. (connector +) Noun phrase with of-phrase fragment:

Figure 4: Distribution of lexical bundles across structural types

general grammatical patterns of use for verb+complement clause construc-

5 A PRELIMINARY FUNCTIONAL TAXONOMY OF THE

bundle serves dierent functions depending on the context. For example,

Table 3: Functional classication of common lexical bundles across registers

Classroom Textbooks Conversation Academic

II. DISCOURSE ORGANIZERS

Classroom Textbooks Conversation Academic

was going to say **

III. REFERENTIAL EXPRESSIONS

IV. SPECIAL CONVERSATIONAL

5.1 Stance bundles

5.1.1 Epistemic stance bundles

Several lexical bundles in classroom teaching combine epistemic stance with

Boys are more likely to be hyperactive, disruptive, and aggressive in class.

5.1.2 Attitudinal/modality stance bundles

bundle if it truly is a question about the participant's preference, but it is more

5.2 Discourse organizing bundles

5.2.1 Topic introduction/focus bundles

As the preceding example shows, sometimes two four-word bundles occur

Topic introducing bundles with second person pronouns invite student

5.2.2 Topic elaboration/clarication bundles

5.3 Referential bundles

imprecision indicators, specication of attributes, and time/place/text refer-

5.3.1 Identication/focus bundles

In many cases, identication/focus bundles also have a discourse organizing

OK. Uh we create a tri-block for an object of type thread, and there is a

In other cases, identication/focus bundles can be used to introduce a

5.3.2 Imprecision bundles

There are obviously companies that do uh evaluations and things like

5.3.3 Bundles specifying attributes

5.3.4 Time/place/text-deixis bundles

5.4 Register variation in the functional exploitation of lexical

Table 4: Distribution of common lexical bundles across functions

Figure 5: Distribution of lexical bundles across functional categories

combines functional characteristics of both conversation (using stance and

5.5 The relationship between structural and functional

Figure 6: Interaction of structural and functional categories

6 CONCLUSION: THE THEORETICAL STATUS OF LEXICAL

function, lexical bundles dier dramatically from other linguistic features